Before You Poll with LLMs: A Deliberative Diagnostic Framework; Analysis Code and Experiment Logs

Summarise with VIRAT
Authors: Ahmed Wali
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Analysis code, experiment output CSVs, and statistical scripts for "Before You Poll with LLMs: A Deliberative Diagnostic Framework." Contains: mixed-effects model scripts (ra4.py), per-persona reversal analysis, threshold sensitivity analysis, and aggregated results for all 22 experimental conditions. Human ground truth data (A1R) is available separately via Harvard Dataverse (doi:10.7910/DVN/ERXBAB, CC0 license) and is not included here.

Fields of Study & Subjects

Computational and Text Analysis Methods Ethics and Social Impacts of AI Opinion Dynamics and Social Influence

Publication details

Journal / source
OSF Preprints (OSF Preprints)
Publication date
17 March 2026
Language
EN
Document type
Preprint